Center Vortices, Nexuses, and the Georgi-Glashow Model John M. Cornwall* Department of Physics and Astronomy 9 University of California, Los Angeles 9 Los Angeles CA 90095 9 1 Abstract n a J In a gauge theory with no Higgs fields the mechanism for confinement is 1 1 by center vortices, but in theories with adjoint Higgs fields and generic sym- metry breaking, such as the Georgi-Glashow model, Polyakov showed that 1 in d=3 confinement arises via a condensate of ’t Hooft-Polyakov monopoles. v 9 We study the connection in d=3 between pure-gauge theory and the theory 3 with adjoint Higgs by varying the Higgs VEV v. As one lowers v from the 0 1 Polyakov semiclassical regime v ≫ g (g is the gauge coupling) toward zero, 0 where the unbroken theory lies, one encounters effects associated with the 9 9 unbroken theory at a finite value v ≃ g, where dynamical mass generation h/ of a gauge-symmetric gauge-boson mass m ≃ g2 takes place, in addition to t the Higgs-generated non-symmetric mass M ≃ vg. This dynamical mass - p generation is forced by the infrared instability (in both 3 and 4 dimensions) e h of the pure-gauge theory. We construct solitonic configurations of the theory : v with both m,M 6= 0 which are generically closed loops consisting of nexuses i X (a class of soliton recently studied for the pure-gauge theory), each paired r with an antinexus, sitting like beads on a string of center vortices with vortex a fields always pointing into (out of) a nexus (antinexus); the vortex magnetic fields extend a transverse distance 1/m. An isolated nexus with vortices is continuously deformable from the ’t Hooft-Polyakov (m = 0) monopole to the pure-gauge nexus-vortex complex (M = 0). In the pure-gauge M = 0 limit the homotopy Π (SU(2)/U(1)) = Z (or its analog for SU(N)) of the 2 2 ’t Hooft-Polyakov monopoles is no longer applicable, and is replaced by the center-vortex homotopy Π (SU(N)/Z ) = Z of the center vortices. 1 N N *E-mail address: [email protected] UCLA/98/TEP/37 December, 1998 1 I. INTRODUCTION InthispaperwedemonstrateasmoothtransitionfromtheGeorgi-Glashow model in the semiclassical limit, where confinement was argued long ago to be due to a condensate of essentially Abelian ’t Hooft-Polyakov monopoles, to the center-vortex picture of confinement as proposed for the pure-gauge theory with no Higgs symmetry breaking. The issues raised are also relevant for an understanding of claims for understanding confinement by Abelian projection. In the seventies several mechanisms for gauge-theory confinement were put forth. The first continuum mechanism to be carefully worked out was Polyakov’s treatment[1] of the d=3 Georgi-Glashow model. He showed that ’t Hooft-Polyakov (TP) monopoles, associated with the breaking of SU(2) to U(1) by an adjoint Higgs field, condensed and confined as would be expected in the much-discussed dual superconductor picture[2]. (To avoid confusion, wenotethatessentiallyAbelianthickvorticesareinvokednotonlyinthedual superconductivity picture, but also in the center-vortex picture put forward here. These are far from being the same; in the center-vortex picture the vor- tices are of magnetic character, but are electric in the dual-superconductor picture. For more modern references to the dual superconductivity hypoth- esis, see, e.g., Ref. [3].) Soon thereafter the center vortex picture[4, 5, 6, 7, 8] of confinement was put forth, based on the idea that a pure-gauge (i.e., no Higgs fields to break the gauge symmetry) theory possessed a kind of quantum soliton which was a fat object of co-dimension two (in particular, a closed string in d=3) car- rying magnetic flux quantized in the center of the gauge group. Much of this work was lattice-oriented, but Ref.[4], working in the continuum, ar- gued that a dynamical gauge-boson mass was generated because of infrared- instability effects, and showed that the effective lagrangian describing this mass (a gauged non-linear sigma model) had Nielsen-Olesen-like vortices. The mass associated with the center vortices is not associated with gauge symmetry breaking; instead, it arises[9] as a necessary element of solving the infrared-unstable Schwinger-Dyson equations of the gauge theory. All N2−1 gauge bosons of SU(N) acquire the same mass m. These vortices could link with Wilson loops and for fundamental Wilson loops whose size scales were large compared to 1/m (hereafter, large Wilson loops) led to topological confinement in which the vortices gave rise to a Wilson-loop phase factor of 2 the form of an element of the center raised to a power which was a linking number: exp(2πiJK/N). Here the integer J specifies the quantized vortex flux and K is the Gauss linking number of the vortex and the Wilson loop. Averaging over these phase fluctuations then led to an area law[4]. It has been shown by lattice-theoretic arguments[10, 11] that in pure- gauge SU(2) only center vortices can confine, by constructing lattice actions in which by adjusting parameters it is possible to retain or exclude thick center vortices and other phenomena. Those actions with no thick center vortices areproved nottoconfine, foranyfinitelatticespacing however small. (Thin vortices confine, but in the small-lattice-spacing limit their action is so large that they are suppressed.) More recently, the center-vortex picture has been revived and various groups[12, 13] have made lattice calculations comparing the area law as com- puted conventionally with the area law computed in various ways. All the ways are related to, but not identical to, the continuum phase approximation which the author has used[4]. This approximation consists, for a given gauge configuration,ofreplacingthetrueWilsonloopvaluebyaphasefactorchosen tobetheelementofthecenternearesttothetruephasefactor(i.e.,forSU(2) one replaces the Wilson loop by its sign). In the continuum it is clear that the phase approximation leaves out perimeter-law terms and short-distance contributions. These lattice calculations show that the phase approxima- tion exactly reproduces the full area law, but their interpretation depends markedly on exactly what version of a phase approximation one uses on the lattice. Kov´acs and Tomboulis[12] have studied the center-vortex picture of confinement both for SU(2) and for SU(3), in both cases finding excellent agreement between the fundamental area law in their phase approximation and the conventional full Wilson-loop calculation. These authors distinguish the behavior of thick vortices (those which, in the continuum, are the ones we discuss here, with a thickness of the inverse physical mass scale) and thin vortices (one lattice spacing thick) by a cooling procedure which destroys the lattice-scale thin vortices, which cannot survive to the continuum limit. Not only do they find that the fundamental-loop area law is exactly reproduced by their phase approximation, they find, as expected, differences at short distances. On the other hand, Greensite and collaborators[13] use different phase approximations, in some cases not fixing a gauge and in some case fixing a gauge. Their essential SU(2) phase approximation is to replace the full 3 fundamental Wilson loop value for a given configuration by its sign. For the fundamental-loop area law they find perfect agreement, in either case, between the full lattice calculations and their approximations. However, when they do not fix a gauge (so-called maximal center projection), the agreement extends to short distances as well. It has been claimed[14, 15, 16] on the basis of an SU(2) character expansion that such agreement is an inevitable consequence, given certain very plausible behavior of higher-J Wilson loops. They thenclaimthat fixing thegaugetothe so-calledmaximal center gauge is, in fact, a meaningful test of the center-vortex picture. The argument is that the gauge fixing is a global construct which can single out thick vortices, while the phase approximations without gauge fixing are infected by lattice-scale vortices. It would take another paper as long as this one now is to discuss these issues thoroughly. We will make only two comments. The first is that while it might be true that the expectation value hZi of the sign Z of the funda- mental Wilson loop W may essentially be hW i itself, this in itself does not F F answer the interesting physical questions connected with the center-vortex picture. One such question is why hZi yields an area law at all (aside from lattice empirics). It could have, for instance, yielded a perimeter law. In fact, an area law for hZi [4, 17, 18] comes about because center vortices have co-dimension two, that is, they are characterized by a two-dimensional density ρ, with the area-law coefficient (string tension) proportional to ρ.1 Another question, hard to address with conventional Creutz-ratio calcula- tions of string tensions, is the actual magnitude of the higher-J Wilson loops invoked in their character expansion. Work is underway in the continuum center-vortex picture to study such loops, but we will not discuss it here. Second, it should be noted that the groups who argue for the trivial equality of the phase approximation and of the full fundamental Wilson loop VEV when no gauge fixing is used have been motivated in part by devel- opments in so-called maximal Abelian projection, or MAP. MAP is an idea introduced long ago by ’t Hooft[21]. He proposed a special way of looking at a gauge theory, by choosing a gauge in which the gauge potentials were Abelian as nearly as possible. Points where the eigenvalues of the gauge potential had degeneracies had to be associated with monopoles, as ’t Hooft 1Evidenceforscalingbehaviorofanarealdensityforvorticesonthelatticeispresented in Refs. [19, 20]. 4 showed. The ’t Hooft gauge fixing was equivalent to breaking the gauge sym- metry(SU(N) → U(1)N−1)withanadjointHiggsfieldofgenericexpectation value, and is relevant to our discuss of the Georgi-Glashow model. Recent lattice calculations[22] are claimed to show that confinement via monopoles can indeed be seen on the lattice by projecting gauge configurations onto the Cartan subalgebra. This projection can be done without gauge fixing, and the same groups[14, 15, 16] who have been concerned with center-vortex gauge fixing have argued that Abelian projection without gauge fixing also does not lead to any significant test of whether ’t Hooft’s MAP monopoles are involved in confinement. They have argued that the MAP-projected the- ory is derived from the non-Abelian theory, and not the other way around. The present paper gives evidence for this point of view not for MAP, but for the transition from Polyakov-like confinement to center-vortex confinement in the Georgi-Glashow model. The above-cited authors also claim that pro- jecting ontoanAbelianensemble isnot important; otherprojections couldbe used, and argue that Abelian projection is essentially trivial, that projection itself has nothing to do with the Abelian dominance claimed to be revealed by projection, and that things are very different with gauge-fixing; only with gauge-fixing can one identify the true physical objects (such as center vor- tices) responsible for confinement. However, one certainly cannot base one’s ultimate understanding of confinement on gauge-dependent properties. The arguments we give here are independent of a choice of gauge. Once again, we forego further detailed comment on these issues, except to notethatMAPisoftendonewithasubsidiarycalculationwhichminimizesan action which is essentially the adjoint-Higgs gauge-boson mass term. Clearly MAP leads to a description of a pure-gauge theory as if it were a Georgi- Glashow model. There is one difference, however: The gauge theory quite independently of any MAP considerations generates a dynamical mass m. A MAP projection may imitate the further generation of a Higgs-mechanism mass M, different from m, as well. So MAP pictures, in general, call for consideration of the Georgi-Glashow model and its monopoles. One must ask whether it is really some form of essentially Abelian monopole or some form of center vortex which truly underlies confinement in either a pure- gauge theory or in the Georgi-Glashow model. This paper argues that it is the center vortex and its near relations which are essential; in this, we agree with Refs. [14, 16]. In particular, we claim that what replaces the TP monopole for finite 5 Figure 1: Schematic picture of a nexus-vortex combination in SU(2). A nexus and an anti-nexus are shown as black circles. They are joined by oppositely-directed vortex segments. m is the combination of nexuses[4, 23, 24] with segments of center vortices, formed into closed loops. These closed loops lead to confinement just as pure center-vortex loops do. Fig. 1 shows a schematic model of a nexus-antinexus pair, connecting regions of center vortex with oppositely-directed fields.2 In actuality the fields extend a distance of order 1/m transverse to the main field direction, indicated by the lines in the figure. This closed loop can be interpreted as a monopole-antimonopole pair with field lines squeezed into tubes, or alternatively it can be interpreted as a center vortex with a nexus-antinexus pair (black circles) on it[23]. Nexuses areconfigurationsinherenttoapure-gaugetheory,andwewillshowthatthey also exist in the Georgi-Glashow model, with its two different mass scales m,M. In this paper we show that a nexus is the essential interpolating element between the Georgi-Glashow model in the semiclassical limit and the pure-gauge theory, where the Higgs VEV v vanishes. Generically, a nexus is a place where up to N center vortices can meet, provided that their flux adds to zero (mod N), for gauge group SU(N). The concept of the nexus (not by that name, however) was introduced long ago[4, 24], but the first quantitative developments came only recently[23]. Quite independently, 2We will be more specific below what we mean by oppositely-directed fields in a non- Abeliangaugetheory. Inanycase,itisclearwhatwemeanwhenwespeakofconventional photonic fields in the Georgi-Glashowmodel. 6 Ambjørn and Greensite[14] have argued in favor of such configurations in the Georgi-Glashowvacuum, andhavegivenacogentdiscussion ofthedifferences between center vortices, the Georgi-Glashow model, and compact QED in d=3. This picture contains, but is certainly not implied by, the picture of con- finement developed by Polyakov[1] for the Georgi-Glashow model in d=3. This model, for gauge group SU(2), identifies two gauge bosons as charged, and the third as the uncharged photon, massless at the classical level. The charged bosons pick up a mass M = vg from the Higgs effect (v is the Higgs VEV). Working in the semiclassical limit where v ≫ g, he showed that there was a condensate of TP monopoles which confined as would a dual superconductor. He further showed that because of quantum effects the condensate density, the string tension, and an induced photon mass m were all exponentially-small in the TP monpole action, which scales like v/g ≫ 1. Naturally, one might expect that in the limit where the charged mass M and the photon mass m were the same, that is, the pure-gauge limit where the Higgs field VEV is zero and there is no symmetry breaking, that center vortices are the mechanism of confinement. Our claim is that in compar- ing the d=3 Georgi-Glashow model and confinement in a pure-gauge theory, the master mechanism of confinement follows from center vortices, and that TP monopoles as they appear in the Polyakov[1] condensate are to be un- derstood as particular cases of the general nexus-vortex configurations we expose here. Once the TP monopoles condense their photonic fields have a mass m which is, in the semiclassical limit, small compared to the charged mass M. Nevertheless, for a (fundamental) Wilson loop whose size scales are large compared to 1/m there is an area law of precisely the type prescribed by the center-vortex picture, following from a linkage of the nexus-vortex combination of Fig. 1 to the Wilson loop, as described above. The flux of a single center-vortex line is half the TP monopole flux, as would be inferred from Fig. 1 by interpreting the nexus as a TP monopole. It therefore should be possible to trace the evolution of the Georgi- Glashow model into the pure-gauge theory by varying the two masses of the theory. There is one mass m for the photon, or third component, of the gauge potential, and another mass M due to symmetry breaking by a Higgs VEV. It is useful to think of m as not just a photonic mass, but as a sym- metric mass present for all three gauge potentials; there is no real difference in the semiclassical regime where m ≪ M. There one has the Polyakov pic- 7 ture, as described above. When M ≪ m one has (essentially) the pure-gauge theory with its center vortices sustained by dynamical mass generation. The two masses can be adjusted by varying the Higgs VEV v from much larger than g toward zero, where there is no symmetry breaking and the pure-gauge theory emerges, except for an unimportant coupling to the massless scalars. But something perhaps unexpected arises: Before reaching v = 0 the Georgi- Glashow model takes on the character of the pure-gauge theory, at a critical value v = v ≃ g. At this point infrared instability of the pure-gauge theory, c in d=3,4, forces the photon mass to become of order g2, the same order as the charged mass M = vg becomes at the critical value. So the Georgi- Glashow model is in the same class as the pure-gauge theory even before the symmetry breaking is restored. In Section II we discuss this consequence of infrared instability, and give a one-loop estimate of the critical value v . At c this point there is essentially only one mass scale (even though m 6= M), and this scale g2 gives the (inverse) distance scale for the transverse extension of the magnetic fields of Fig. 1, and there is no qualitative difference between center-vortex confinement and confinement in the Georgi-Glashow model. Naively it might appear that the configurations of Fig. 1 could easily be understood in some Abelian version of the theory, just as center vortices themselves and the TP monopole have a certain Abelian character. But this is wrong; any configuration of gauge fields which has a non-zero magnetic flux over a sphere at infinity is necessarily non-Abelian, as evidenced by the homotopy Π (SU(2)/U(1)) = Z . In Section III we discuss the non-trivial 2 2 constructions which lead to a qualitative description of a quantum soliton depending on the two scales m,M and capable of describing a TP monopole at m = 0 and a pure-gauge nexus-vortex combination at M = 0. We do not give a quantitative treatment of the soliton, which will be deferred to later work. II. INFRARED INSTABILITY IN THE GEORGI-GLASHOW MODEL The Georgi-Glashow model is a Yang-Mills theory coupled to an adjoint Higgsfield. ItcanbedefinedforanySU(N), andwithgenericVEVsitbreaks this symmetry down to U(1)N−1. We will only consider it explicitly for the originally-proposed model where the gauge group SU(2) is broken to U(1), andwewillonlyworkoutthed=3case(asPolyakov[1]did). Twoofthegauge 8 bosons (carrying U(1) charge, identified with electromagnetic (EM) charge) acquire a mass, while the third, the photon, remains massless. This model has TP monopoles with long-range EM magnetic fields; asymptotically, the monopole fields are precisely those of the Wu-Yang singular monopoles of the corresponding pure-gauge theory. The action for this model is: 1 1 λ S = d3x{ (Ga)2 + (D ψa)2 + [(ψa)2 −v2)]2} (1) Z 4g2 ij 2 i 8 We will often use the conventional antihermitean matrices τa τa A = Aa, ψ = ψa, D = ∂ +A . (2) i 2i i 2i i i i The VEV of, say, ψ3 is v; the mass M of the charged gauge bosons is M = gv and the mass M of the Higgs particle is M = (λ)1/2v. Mostly we are H H interested in the λ → ∞ limit, where the massive Higgs particle decouples (but not the Goldstone fields). The semiclassical limit of the theory is v ≫ g, or M ≫ g2. The action of the TP monopole is then large: 4πζM 4πζv S = = ≫ 1, (3) TP g2 g where ζ is a numerical constant of order unity. In the semiclassical theory Polyakov shows that there is a condensate of TP monopoles with a density proportional to exp(−S ) which is exponentially-small in v/g. The string TP tension is proportional to this density and therefore is exponentially-small, and the condensate causes the TP monopoles to acquire a mass m which is exponentially-small too. For all practical purposes this small mass m can be ignored. How can one go from the Georgi-Glashow model to the pure-gauge the- ory, with no Higgs fields? At the classical level, to decouple the scalar fields requires changing the sign of v2 in the action (1), thereby removing the Goldstone fields which give the charged gauge bosons their mass. If v2 turns negative the symmetry is restored, and all the particles of the scalar sec- tor acquire the same mass. Ultimately the scalar sector can be effectively decoupled by making that mass large enough. 9 We will study the transition between the semiclassical Higgs regime and the pure-gauge regime by reducing v toward zero from a value much larger than g. Clearly, at v = 0 the symmetry breaking is turned off, and one has a gaugetheorycoupled tothreemassless scalar fields (thisisnot, aswewill see, animportant coupling). At first it may appear that even when v ≃ g the the- ory looks much like the Abelian-monopole phase, with long-range EM fields for an isolated TP monopole and massive charged gauge bosons. However, this is not so. Because of the underlying infrared instability of the pure- gauge theory[17, 30, 18] when the charged mass M is small enough, tachyons appear in the S-matrix as calculated in one-loop3 perturbation theory. The cure for these tachyons[9, 26] is a dynamically-generated mass, having noth- ing to do with Higgs effects, which must be large enough to overcome the tachyonic instability. Another example of the same phenomenon occurs in Yang-Mills-Chern-Simons (YMCS) theory, where the CS term produces a mass of classical value kg2/4π at level k. However, if k is less than a critical value of order[30] 2N in SU(N), the tachyon persists, at least at one-loop level, and dynamical mass generation must take place. This dynamical mass is of order Ng2. The main technique for uncovering these results is the pinch technique (PT)[9, 31, 32, 33]. In the PT a gauge-invariant gauge-boson propagator is extracted from the S-matrix by incorporating pieces of vertex, box, and other graphs which have the kinematic structure of propagator parts into the usual propagator defined by Feynman graphs. Since the S-matrix is gauge-invariant, so is the resulting propagator. This propagator-like kine- maticstructurearises frompinching outcertainlines inthese non-propagator graphs by elementary applications of Ward identities. Although we will be more precise momentarily, it is useful to indicate crudely what is going on. Roughly, the structure of the d=3 Euclidean PT propagator (denoted with a hat), when some of the gauge bosons pick up a mass M = vg from the Higgs field, is (omitting inessential longitudinal terms 3Is a one-loop result even qualitatively right? There is some evidence that it is, from Eberlein’s paper[29] where he calculates two-loop results which are quite close to previously-calculated one-loop gap-equation results for the d=3 gauge boson mass. How- ever, most of these one-loop gap equation results are infected with tachyons[18], coming from calculated mass values which are too small to cure the infrared instability. 10